boost timeout slightly for connections and add another call to wait for a stable cluster before killing a node